Skip to content

Quick start in 5 minutes

This guide shows how to get Rabetbase CLI 2.0 up and running quickly, in both new and existing projects.

If you already use an AI coding tool such as Claude Code, Codex, Cursor, Trae, or Windsurf, just copy the prompt below to your AI assistant. Following the guide, it will install the CLI, install the Skill, and complete the login — you only step in for browser authorization or project confirmation.

Prompt (copy it straight to your AI assistant):

Mainland China node:

帮我安装 Rabetbase CLI:https://www.lovrabet.com/agent-install-guides/cn/rabetbase-cli.md

Indonesia node:

帮我安装 Rabetbase CLI:https://www.lovrabet.id/agent-install-guides/id/rabetbase-cli.md

Notes:

INFO

  1. Authorization is where AI Agent installs most often get stuck: when the Agent pastes the actual Login URL, open the link and complete the browser authorization first, then return to the chat and tell the Agent "authorization complete".
  2. If the Agent asks whether to initialize a project, confirm only when the current directory really is the target project. If you're just installing the CLI, there's no need to run rabetbase project init.

Supported AI Agent tools: Claude Code, Codex, Cursor, Qoder, Trae, Windsurf, GitHub Copilot, OpenCode, Hermes Agent, OpenClaw

💡 If you don't use AI Agent tools, or you want to control every step yourself, continue with the manual installation below.


Manual installation

For: first-time Rabetbase CLI users whose machines have never had any version installed.

>> Watch the installation screencast

1.1 Install the CLI

Package: @lovrabet/rabetbase-cli

bash
# 全局安装(推荐)
npm install -g @lovrabet/rabetbase-cli@latest

# 验证安装
rabetbase --version

1.2 Install the Rabetbase Skill

Use the CLI's built-in entry point:

bash
rabetbase skill install

If you need the raw skills command, this works too:

bash
npx skills add lovrabet/rabetbase -g -y

Install it globally so your AI assistant can use rabetbase from any project directory.

1.3 Authenticate

bash
rabetbase auth login

Complete the OAuth login in the browser that opens; the CLI saves the session automatically. One login covers all projects.

1.4 Initialize a project

Initialize configuration inside an existing project directory:

bash
cd your-project/
rabetbase project init

Create a new project from scratch:

bash
rabetbase project create

1.5 Quick verification

bash
rabetbase doctor
rabetbase schema --format compress
rabetbase app list --format compress

To confirm --jq is available, run one more:

bash
rabetbase schema --format compress --jq '.data.cli.version'

2. Changes worth knowing in the current release

  • --jq is more stable: the latest release bundles a jq sidecar, so it no longer relies on node-jq downloading a binary from GitHub at install time — better stability on mainland China networks and in CI.
  • Dataset listings are more complete by default: rabetbase dataset list now returns DO V2 datasets by default, including DB_TABLE and METADATA. Add --source DB_TABLE to see database tables only, or --source METADATA to see metadata only.
  • The AppCode environment variable must be used explicitly: CI pipelines can still store RABETBASE_APPCODE, but commands must pass --appcode "$RABETBASE_APPCODE". The CLI no longer silently overrides project configuration from an environment variable.
  • Writes are safer by default: rabetbase page generate-start returns only a dry-run preview unless you explicitly add --apply, and writing project configuration no longer drags global app settings into project files.
  • Dataset write operations require an explicit target: dataset rename, field-update, extend-update, delete, and similar commands no longer rely on fuzzy context. Select the target with --code, --id, or --dbid, and run --dry-run first.
  • Menu asset updates are more conservative: when rabetbase menu update refreshes CDN assets, it preserves the page's existing extend configuration, reducing the risk of overwriting production settings.
  • You can now query real data during development: rabetbase data filter and rabetbase data getOne are official commands — handy for quickly checking what real data looks like at dev time.
  • One unified Skill install entry point: just run rabetbase skill install — no need to memorize the underlying npx skills add ... form.

3. Checklist

Fresh install

  • [ ] Install the CLI: npm install -g @lovrabet/rabetbase-cli@latest

  • [ ] Install the Skill: rabetbase skill install

  • [ ] Authenticate: rabetbase auth login

  • [ ] Initialize a project: rabetbase project init or rabetbase project create

  • [ ] Verify: rabetbase doctor

4. Next steps

  • Using Rabetbase in Claude Code: from one sentence to a full feature
  • Configuration: apps, environments, multiple apps
  • Rabetbase CLI changelog

基于飞书知识库同步生成,内容以飞书源文档为准